Generative Grammar

Theory and its History

Robert Freidin

• Editorial: Routledge
• Año: 2007
• ISBN:
  9781134322107
• Edición: 1
• Categoría: Languages & Linguistics
• Categoría: Linguistics

Generative Grammar presents a substantial contribution to the field of linguistics in drawing together for the first time the author's most significant work on the theory of generative grammar.

The essays collected here display Freidin's role in moving the theory forward in terms of new proposals, and analyse the efforts to understand the evolution and history of the theory by careful investigation of how and why it has changed over the years.

1
Introduction

The biolinguistic approach to the study of human language is a fairly recent development in the history of linguistics, a perspective that developed within modern generative grammar (from LSLT to the present). This approach seeks to understand what presumably unique biological properties account for human language, more specifically its structure, use and biological origin. It has been pursued by postulating explicit computational models of what a speaker of a human language must know to be able to use the language, the knowledge of linguistic structure that underlies linguistic performance. Such models have been the subject of study in a subfield of linguistics called generative grammar.

The papers in this volume deal with central topics within the study of generative grammar—primarily, the theories of movement, case and binding, as well as their intersections and empirical motivation. They also cover the broader history of the field, which is rich and intricate. This history provides a context for a fuller understanding of current proposals, which after all also form an integral part of this history. Thus the separation of these essays into two parts, theory and history, is somewhat artificial. Current theoretical discussions simply contribute to the ongoing history and often provide important clarifications of previous work. Historical discussions usually clarify the past and often create a context in which to understand what progress, if any, has been or is being made. Furthermore, ideas that have been abandoned along the way can be resurrected and refurbished in the current context—the revival of generalized transformations being a spectacular example.

A Theory

The syntactic cycle has played a central role in the theory of movement since its inception in Chomsky 1965. “Cyclicity and the theory of grammar” (1978) (chapter 2) resulted from research on how the cycle (in particular the Strict Cycle Condition of Chomsky 1973 (henceforth SCC)) operated under the trace theory of movement rules. The empirical motivation for the SCC discussed in Chomsky 1973 cited one example of what is commonly referred to as a wh-island violation. One derivation of this example violated no known movement constraints and yet yielded deviant output. Thus the SCC appeared to be necessary to rule out this derivation. Under trace theory however, this derivation yielded the same output as a derivation that violated some other constraints on movement (e.g., the Subjacency Condition). While the SCC could only be interpreted as a condition on derivations, Subjacency could also be interpreted as a condition on representations—more precisely, a locality condition on trace binding. Under this interpretation, the SCC becomes superfluous. “Cyclicity and the Theory of Grammar” generalized this result to other cases, not considered in Chomsky 1973, whose derivations also violated the SCC. It demonstrated how most wh-movement cases could be handled by Subjacency¹, whereas NP-movement cases involved other conditions.² By deriving the empirical results of the principle of the syntactic cycle from other independently motivated general principles of grammar, this paper demonstrated how the cyclicity of the computational system was in fact built into the general architecture of UG. This result also raised the issue of the derivational vs. representational interpretation of general principles and provided an argument for the latter interpretation, given that some violations of the SCC also violated Subjacency interpreted as a condition on representations.

Like the previous chapter, chapter 3, “Superiority, subjacency, and economy” (1995), concerns the potential for overlap among conditions—the two mentioned in the title and Chomsky’s Shortest Movement economy constraint—as applied to the derivation of certain wh-island violations. For example, constructions like (1) might be derived in two different ways, one of which violates both the Subjacency and Superiority conditions while the other only violates the latter, depending on whether who moves to Spec-CP in the complement clause.

(1) *What did you forget who had borrowed?

This paper attempts to refine the analysis of wh-island violations by using wh-phrases of the form which-N instead of bare interrogative pronouns. The Superiority Condition does not apply to movement involving which-N phrases in single clauses, as illustrated by the well-formedness of examples like (2).

(2) Which books did which students borrow?

This result extends to such wh-phrases in complex sentences. Thus (3) in contrast to (1) appears to be relatively well-formed.

(3) Which books did you forget which students had borrowed?

To account for the contrast between (1) and (3) within a minimalist feature checking analysis circa 1994 this paper adopts a rather radical stance, including the Form Chain analysis of Chomsky 1993, countercyclic movement (which reinforces the conclusion of the previous chapter that the SCC cannot be a primitive, but rather its appropriate empirical effects are derived from other principles of grammar), and the rejection of the Shortest Move condition, which ought to block (3) and possibly (2) as well.³

Between the publication of chapter 2 in 1978 and the publication of “Cyclicity and minimalism” (chapter 4) in 1999, the theory of phrase structure underwent a radical revision. Starting in 1979 it became clear that PS rules redundantly stipulated properties that followed from the interaction of general principles of grammar (e.g., the Case Filter and the θ-Criterion) and idiosyncratic properties of individual lexical items (Chomsky class lectures 1979; see also Stowell 1981). Thus phrase structure rules were abandoned as the mechanism for building phrase structure. However, it took over a decade before the notion of generalized transformation was revived as the mechanism for constructing phrase structure from lexical items (Chomsky 1993, 1995b).⁴

With PS rules, derivations of phrase structure were purely top-down, whereas with Merge, these derivations are exclusively bottom-up. This fundamental difference prompted a reevaluation of how cyclic derivation worked as well as the empirical motivation for a cyclic principle and the possibility of deriving its empirical effects from other independent principles of grammar, the major topics of chapter 4.

It is interesting to note that a cyclic principle (e.g., the SCC) ensures that the phrase marker of sentence will be processed strictly bottom-up (i.e., from smaller to larger domains) even (or especially) when phrase structure is constructed top-down. Under minimalist analysis the strictly bottom-up creation of phrase structure is determined by an extension condition (Chomsky 1993), which requires that each step of a derivation extends the right or left edge of the phrase marker affected.⁵ This applies to movement operations as well, and thereby serves as a replacement cyclic principle (to the SCC) without mentioning the notion of cyclic domain. In effect, Chomsky’s Extension Condition (1993, p. 22) provides another way to derive the empirical effects of a cyclic principle.

Under minimalism the details of derivations involving multiple movements (e.g., the derivation of the wh-island violations covered in the previous two chapters) are both more complicated and less well determined by the theory of grammar because of the wide choice of analytic options. Is movement the result of Move or Attract? Do some features move independently of the rest of the constituent they occur in, which then undergoes displacement by some form of generalized pied piping (Chomsky 1995d) or do whole constituents move to check features? Is there a strong/weak feature distinction in addition to the interpretable/uninterpretable distinction? And if so, does an unchecked strong feature cancel a derivation—an analysis that leads to a further, perhaps questionable, distinction between deletion and erasure?

Chapter 4 attempts to sort through these options for constructions whose derivations constitute SCC violations—including super-raising, the Head Movement Constraint (HMC, see Travis 1984) and Constraint on Extraction Domains (CED, see Huang 1982) violations, as well as wh-islands. It shows how, under current minimalist analysis, the empirical evidence for cyclic derivation follows from other independently motivated grammatical principles, and thus eliminates the need to stipulate an independent cyclic principle. For example, in the case of wh-island violations, if [+wh] is a strong feature of C that motivates movement to Spec-CP, then if the feature is not checked immediately after it enters the derivation, the derivation cancels. If the feature is checked immediately after it enters the derivation, then it is no longer active and hence cannot motivate a further counter-cyclic movement of a wh-phrase. In this case the empirical effects of cyclicity follow from the principles of feature checking and so there is no need to invoke an independent cyclic principle.⁶

Chapter 4 also discusses three additional proposals for deriving cyclicity from other principles of grammar. Kitahara 1995 proposes that the economy condition requiring shortest derivations always blocks a countercyclic movement because it involves an extra derivational step. Collins 1997 suggests that countercyclic movements result in configurations that violate the Linear Correspondence Axiom of Kayne 1994. Chapter 4 identifies potential flaws in these proposals and proposes instead that countercyclic Merge might be ruled out because the elementary operation that performs merger is incapable by its nature of applying countercyclically. Thus no condition on derivations or on representations is needed to block countercyclic operations. The formulation of the elementary operation suffices, an optimal solution on minimalist assumptions.

Like the analysis of cycilicity the analysis of grammatical Case has played a fundamental role in a theory of syntactic movement. The theoretical importance of Case for modern generative grammar was first spelled out in an unpublished letter by Jean-Roger Vergnaud to Chomsky and Lasnik about their paper “Filters and Control” (1977).⁷ Chomsky 1980 adapts Vergnaud’s Case theory in a formulation of a Case filter, which limits the distribution of NPs with phonetic content (as opposed to phonetically empty NPs—e.g., trace and PRO) to Casemarked positions. The Case filter analysis provides a more general and more principled account of the distribution of phonetically realized NPs, as well as a principled motivation for the movement of nominal expressions in general. Thus, NPs with phonetic content that enter a derivation in a Caseless position must move to a Casemarked position to yield a legitimate syntactic construction.

Chapter 5 (“Core Grammar, Case Theory, and Markedness” (1981)), an early study of the new Case theory, works out some ramifications of the Case filter analysis. In particular, it is concerned with the interaction (and hence the ordering) of Case-assignment and Deletion, and also of the Case Filter and Deletion. The paper identifies empirical evidence that determines how these mechanisms must be ordered. It also demonstrates how Case theory distinguishes between NP-trace and wh-trace, where only the latter is subject to the Case Filter. This distinction is further supported by Binding Theory, as discussed in chapter 8. This result raised a question about the nature of the Case Filter, which had been assumed to apply only to phonetically realized NPs. The inclusion of wh-trace, which is obviously not phonetically realized, suggests that the Case filter analysis may not be properly formulated. This concern led to a “visibility” approach to Case (Chomsky 1981, 1986) that integrates Case theory and θ-theory This approach attempts to explain the evidence from wh-movement as a violation of the θ-Criterion rather than the Case filter.⁸

Initially Case theory was formulated primarily on the basis of English, a language without a rich morphological Case system. Expanding Case analysis to languages that have rich morphological Case systems (e.g., Russian and Icelandic) revealed a further general principle as well as some refinements of Case theory. Such languages usually manifest two distinct types of morphological Case: configurational and lexical (a.k.a. quirky Case). Configurational Case is assigned purely in terms of syntactic position, whereas lexical Case is assigned via selection by a specific lexical head (where different heads of the same category may select different lexical Cases). In constructions where configurational and lexical Case could be in conflict (e.g., the object of a verb that assigns lexical Case), the lexical Case assignment must be satisfied and therefore the configurational Case is morphologically suppressed. This follows from the principle of Lexical Satisfaction of Freidin & Babby 1984, the ramifications of which are investigated in chapter 6, “Lexical case phenomena” (1991). Furthermore, lexical Case phenomena establish a distinction between Case assignment and Case licensing. In a clause whose main verb selects a lexically Casemarked subject, that lexically Casemarked subject must occur in a position that is configurationally licensed for Case. It is necessary but not sufficient that the subject bears the appropriate lexical Case. Thus a phonetically realized NP must be Case licensed as well as Casemarked. For configurational Case, licensing and Casemarking appear to be indistinguishable, but for lexical Case these are distinct processes.

Under minimalism, the Case Filter has been replaced by the Principle of Full Interpretation (FI), which has subsumed its empirical effects.⁹ This follows given that all phonetically realized NPs enter a derivation with unvalued Case features and that because these features are uninterpretable at PF and LF (with or without values), they must be eliminated via checking during the course of the derivation. If not, these unchecked features violate FI at PF and LF. It is further assumed that the valuation and checking of Case features is a reflex of the checking of agreement features (henceforth

) Nominative Case is valued and checked via

the of T, and accusative Case via the

of ν.

The role of Case in the theory of movement has also changed significantly under minimalism. In the initial discussions of Case theory it was assumed that NP-movement (e.g., passive and raising) were driven by the need for a NP with phonetic content to be Case-marked. Under minimalism, Case-marking (i.e., valuation of Case features) is a secondary effect, t...

Índice

• Cover Page
• Title Page
• Copyright Page
• Acknowledgements
• 1. Introduction
• Part I: Theory
• Part II: History
• References